Attributable Risk

The attribrisk package has been created to make attributable risk (AR) inference easier to calculate. This vignette gives only a small amount of theory, users who wish more are directed to Khan [3]. The term attributable risk is used to refer to the amount of disease in a population that can be attributed to a particular exposure. The terms population attributable risk and etiologic fraction have also been used to refer to the same concept. Consider the data from Whisnant et al [6] shown in table 1. Assume for the moment that this data were a simple sample of size 2644 from the population at large (which it is not). Then the rate of ischemic stroke among those without high blood pressure would be estimated as 559/943=0.41 while that in those with high blood pressure it is 0.55. The attributatable risk is the fraction of the strokes in the second group that is presumed to be directly due to their high blood pressure, namely (0.55− 0.41)/0.55 = 0.25. In reality the data set above is not a population sample, but a collection of all ischemic stroke patients in Olmsted County over a defined period along with a set of age matched controls, and the proper calculation of AR becomes more complex in that case. Define the following probabilities. They may be interpeted as prevalence rates, incidence rates or proportions depending on the circumstances.